AI chatbots link pregnant users to anti-abortion sites, probe finds

The investigation by AlgorithmWatch tested four major chatbots with three fictional personas dealing with unplanned pregnancies, producing 270 responses in English, German, and Italian. The findings show that AI chatbots, even with built-in search, function as black boxes shaped by training data and other factors that are difficult to audit. The empathetic tone and apparent completeness of the responses make them feel trustworthy, which is a key difference from traditional search engines.

A court in Munich has ruled that AI-generated answers must be treated as original content, meaning the limited liability protections for search engines don't apply. Germany's media regulators agree, classifying AI search engines and chatbots as content providers with full legal responsibility. However, the investigation suggests these obligations would be hard to meet, given problems that may be nearly impossible to fix on a technical level. Google and OpenAI pointed to their policy guidelines, which don't specifically address abortion-related queries, while an OpenAI spokesperson noted that GPT-5.5, shipped after the testing period, is better at taking user context into account.